A dynamic Bayesian network for protein secondary structure prediction. (A) The first column shows the variables of the prologue (models the first amino acid) and the second column shows the variables of the chunk (models the second up to the last amino acid). The chunk is rolled (i.e.) extended to the right to get the final network. (B) An example state sequence and the values of state count down and change state variables for D
= 7. state count down and change state are used to control transitions from one secondary structure segment to the next and model the length distribution of segments.